# Fine-Tuning Language Models with Reward Learning on Policy

arXiv: https://arxiv.org/abs/2403.19279 · HTML: https://mlsanity.com/paper/2403.19279/

## Key facts

- Cited by 15 papers in the map (data through 14 Oct 2026).
- Placed in Reward modeling and RLHF, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 40 of its 53 references that are in the map.
- First posted on arXiv 28 Mar 2024.

## Placed in

- LLM alignment / RLHF / preference optimisation / [Reward modeling and RLHF](https://mlsanity.com/alignment/reward-modeling-and-rlhf/)

## Builds on (most cited first, 30 of 40)

- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Measuring Massive Multitask Language Understanding](https://mlsanity.com/paper/2009.03300/) (2020)
- [Self-Consistency Improves Chain of Thought Reasoning in Language Models](https://mlsanity.com/paper/2203.11171/) (2022)
- [On the Opportunities and Risks of Foundation Models](https://mlsanity.com/paper/2108.07258/) (2021)
- [Deep reinforcement learning from human preferences](https://mlsanity.com/paper/1706.03741/) (2017)
- [Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2204.05862/) (2022)
- [Constitutional AI: Harmlessness from AI Feedback](https://mlsanity.com/paper/2212.08073/) (2022)
- [Scaling Instruction-Finetuned Language Models](https://mlsanity.com/paper/2210.11416/) (2022)
- [Unsupervised Learning of Visual Features by Contrasting Cluster Assignments](https://mlsanity.com/paper/2006.09882/) (2020)
- [Learning to summarize from human feedback](https://mlsanity.com/paper/2009.01325/) (2020)
- [DeBERTa: Decoding-enhanced BERT with Disentangled Attention](https://mlsanity.com/paper/2006.03654/) (2020)
- [Fine-Tuning Language Models from Human Preferences](https://mlsanity.com/paper/1909.08593/) (2019)
- [Language Models (Mostly) Know What They Know](https://mlsanity.com/paper/2207.05221/) (2022)
- [Multitask Prompted Training Enables Zero-Shot Task Generalization](https://mlsanity.com/paper/2110.08207/) (2021)
- [Learning deep representations by mutual information estimation and maximization](https://mlsanity.com/paper/1808.06670/) (2018)
- [Scaling Laws for Reward Model Overoptimization](https://mlsanity.com/paper/2210.10760/) (2022)
- [A General Language Assistant as a Laboratory for Alignment](https://mlsanity.com/paper/2112.00861/) (2021)
- [Teaching Large Language Models to Self-Debug](https://mlsanity.com/paper/2304.05128/) (2023)
- [Super-NaturalInstructions: Generalization via Declarative Instructions on 1600+ NLP Tasks](https://mlsanity.com/paper/2204.07705/) (2022)
- [Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2307.15217/) (2023)
- [Teaching Models to Express Their Uncertainty in Words](https://mlsanity.com/paper/2205.14334/) (2022)
- [AlpacaFarm: A Simulation Framework for Methods that Learn from Human Feedback](https://mlsanity.com/paper/2305.14387/) (2023)
- [The Flan Collection: Designing Data and Methods for Effective Instruction Tuning](https://mlsanity.com/paper/2301.13688/) (2023)
- [Learning Representations by Maximizing Mutual Information Across Views](https://mlsanity.com/paper/1906.00910/) (2019)
- [Can Large Language Models Be an Alternative to Human Evaluations?](https://mlsanity.com/paper/2305.01937/) (2023)
- [RLAIF vs. RLHF: Scaling Reinforcement Learning from Human Feedback with AI Feedback](https://mlsanity.com/paper/2309.00267/) (2023)

## Built on by (most cited first, 15 of 15)

- [On the Limited Generalization Capability of the Implicit Reward Model Induced by Direct Preference Optimization](https://mlsanity.com/paper/2409.03650/) (2024)
- [Real-Time Aligned Reward Model beyond Semantics](https://mlsanity.com/paper/2601.22664/) (2026)
- [ORIGEN: Zero-Shot 3D Orientation Grounding in Text-to-Image Generation](https://mlsanity.com/paper/2503.22194/) (2025)
- [ReDit: Reward Dithering for Improved LLM Policy Optimization](https://mlsanity.com/paper/2506.18631/) (2025)
- [The Perils of Optimizing Learned Reward Functions: Low Training Error Does Not Guarantee Low Regret](https://mlsanity.com/paper/2406.15753/) (2024)
- [Off-Policy Corrected Reward Modeling for Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2507.15507/) (2025)
- [Modeling and Optimizing User Preferences in AI Copilots: A Comprehensive Survey and Taxonomy](https://mlsanity.com/paper/2505.21907/) (2025)
- [Non-differentiable Reward Optimization for Diffusion-based Autonomous Motion Planning](https://mlsanity.com/paper/2507.12977/) (2025)
- [OPPO: Accelerating PPO-based RLHF via Pipeline Overlap](https://mlsanity.com/paper/2509.25762/) (2025)
- [EvoRS: On-Policy Self-Evolution of Reward Systems for Open-Ended Reinforcement Learning](https://mlsanity.com/paper/2609.12459/) (2026)
- [Navigation-GPT: A Robust and Adaptive Framework Utilizing Large Language Models for Navigation Applications](https://mlsanity.com/paper/2502.16402/) (2025)
- [Context-Emotion Aware Therapeutic Dialogue Generation: A Multi-component Reinforcement Learning Approach to Language Models for Mental Health Support](https://mlsanity.com/paper/2511.11884/) (2025)
- [UMM-RM: An Upcycle-and-Merge MoE Reward Model for Mitigating Reward Hacking](https://mlsanity.com/paper/2512.00724/) (2025)
- [Imitation Learning for Elder-Facing Speech Synthesis](https://mlsanity.com/paper/2606.21053/) (2026)
- [Re2A: Situated Conversational Recommendation via Rubric-based Preference Reasoning and Alignment](https://mlsanity.com/paper/2609.18249/) (2026)

## Abstract (arXiv)

Reinforcement learning from human feedback (RLHF) has emerged as an effective approach to aligning large language models (LLMs) to human preferences. RLHF contains three steps, i.e., human preference collecting, reward learning, and policy optimization, which are usually performed serially. Despite its popularity, however, (fixed) reward models may suffer from inaccurate off-distribution, since policy optimization continuously shifts LLMs' data distribution. Repeatedly collecting new preference data from the latest LLMs may alleviate this issue, which unfortunately makes the resulting system more complicated and difficult to optimize. In this paper, we propose reward learning on policy (RLP), an unsupervised framework that refines a reward model using policy samples to keep it on-distribution. Specifically, an unsupervised multi-view learning method is introduced to learn robust representations of policy samples. Meanwhile, a synthetic preference generation approach is developed to simulate high-quality preference data with policy outputs. Extensive experiments on three benchmark datasets show that RLP consistently outperforms the state-of-the-art. Our code is available at urlhttps://github.com/AlibabaResearch/DAMO-ConvAI/tree/main/rlp.